Topic: Avalon recondition (Read 2090 times)

Thank you for this johnyj!
Can you please make measurements of the whole module?





According to me, we did not yet reached the optimal potential of the Avalon hardware, there is still a lot of work to do, Avalon is running for example on p2pool with 30% stale shares, even if that is a miner/mine thing, im sure by optimal cooling, the lifetime of a Avalon device can be prolonged.
And of course there is the question, where the limit of the device is, since we dont get any test data from Avalon related to this questions, we probably have to do that ourselfs.

The difference is about 0.1 mm, thermal compound should be enough to fill the gap

It seems to be so, 23 screws are enough to transfer the heat into the heatsink, I noticed on that thermal photo on Avalon wiki page that the screws showed the lowest temp

Anyway someone has burnt their board and I'd better be careful

Since the contact is so uneven wouldn't it better to go with 0,5/1mm thick thermal pads, or maybe 2mm really soft ones? Would need quite a bit of it though and could get a bit pricy 

This is prob beneficial for people who may want to overclock or optimize hash/joule to maximize profit.

yifu stated several times at the conference that the chips will run fine at 100 C. heat is not an issue.

Certainly looks worthwhile then for every Avalon owner to get some Artic Silver and goop it up.

Incredible how poorly applied the thermal compound was coming out of the factory. Also incredible how the ~200w blade was able to not overheat with such poor thermal transfer. I guess it must have been radiating a lot of heat from the top of the PCB?

edit: One other thing, you might want to put some thermal paste under the VRM section. They aren't producing a lot of heat but it is something to consider.

It's getting hot and I tried recondition of the avalon, took 7 hours!

To my surprise, the whole hash module is one piece of PCB board, and although thermal compound are painted on the back of PCB board, only several spots of the PCB have a good contact with thermal compound on the heatsink, it means the majority of chips have no contact with the heatsink at all and most of the heat transfer job is done by those screws (23 of them) Obviously avalon team do not have enough time to add thermal compound for each individual chip, and those chips actually are not that hot



Further examination of the PCB board shows, those exposed metal pads on the back of the chips are lower than the PCB board surface, they will never get a direct contact with the heatsink





I dropped different amount of MX4 thermal compund on different location of the PCB board to test the contact, without tightening the screw, they all spread out nicely. Applied thermal compund in this way, for all the metal pads, 240 of them...



Broke a tiny capacitor by accident, tried different method, no way to put it back on since the broken trace is too small to accept solder    (Luckily this does not affect hash rate yet)

Result is reduced 1 degree temp and lowered fan rpm, for a fixed fan rpm the effect should be 3 degree lower temp. I am more confident with the health of each chip now

By the way, using an external desktop fan to cool the bottom plate is extremly helpful, can lower the temp by another 2 degree